Feature/ai/hourly predictions

frontend/occupi-mobile4/package.json

@@ -13,18 +13,21 @@
     "lint": "expo lint"
   },
   "jest": {
-  "preset": "jest-expo",
-  "transformIgnorePatterns": [
-    "node_modules/(?!((jest-)?react-native|@react-native(-community)?)|expo(nent)?|@expo(nent)?/.*|@expo-google-fonts/.*|react-navigation|@react-navigation/.*|@unimodules/.*|unimodules|sentry-expo|native-base|react-native-svg|@gluestack-ui/themed|expo-device)"
-  ],
-  "collectCoverage": true,
-  "coverageReporters": ["lcov", "text"],
-  "coverageDirectory": "coverage",
-  "setupFiles": [
-    "./node_modules/react-native-gesture-handler/jestSetup.js",
-    "./jest.setup.js"
-  ]
-},
+    "preset": "jest-expo",
+    "transformIgnorePatterns": [
+      "node_modules/(?!((jest-)?react-native|@react-native(-community)?)|expo(nent)?|@expo(nent)?/.*|@expo-google-fonts/.*|react-navigation|@react-navigation/.*|@unimodules/.*|unimodules|sentry-expo|native-base|react-native-svg|@gluestack-ui/themed|expo-device)"
+    ],
+    "collectCoverage": true,
+    "coverageReporters": [
+      "lcov",
+      "text"
+    ],
+    "coverageDirectory": "coverage",
+    "setupFiles": [
+      "./node_modules/react-native-gesture-handler/jestSetup.js",
+      "./jest.setup.js"
+    ]
+  },
   "dependencies": {
     "@eva-design/eva": "^2.2.0",
     "@expo/vector-icons": "^14.0.0",
@@ -45,19 +48,23 @@
     "@ui-kitten/components": "^5.3.1",
     "axios": "^1.7.2",
     "date-fns": "^3.6.0",
-    "expo": "~51.0.14",
+    "expo": "^51.0.32",
     "expo-blur": "^13.0.2",
+    "expo-checkbox": "~3.0.0",
     "expo-constants": "~16.0.2",
     "expo-device": "~6.0.2",
     "expo-file-system": "^17.0.1",
     "expo-font": "~12.0.7",
+    "expo-haptics": "~13.0.1",
     "expo-linear-gradient": "~13.0.2",
     "expo-linking": "~6.3.1",
     "expo-local-authentication": "^14.0.1",
+    "expo-location": "~17.0.1",
     "expo-notifications": "~0.28.9",
     "expo-print": "~13.0.1",
     "expo-router": "~3.5.16",
     "expo-secure-store": "~13.0.2",
+    "expo-sensors": "~13.0.9",
     "expo-sharing": "~12.0.1",
     "expo-splash-screen": "~0.27.5",
     "expo-status-bar": "~1.12.1",
@@ -89,11 +96,7 @@
     "react-native-web": "~0.19.10",
     "reanimated-color-picker": "^3.0.4",
     "tinycolor2": "^1.6.0",
-    "zod": "^3.23.8",
-    "expo-location": "~17.0.1",
-    "expo-sensors": "~13.0.9",
-    "expo-haptics": "~13.0.1",
-    "expo-checkbox": "~3.0.0"
+    "zod": "^3.23.8"
   },
   "devDependencies": {
     "@babel/core": "^7.25.2",

frontend/occupi-mobile4/screens/Dashboard/Dashboard.tsx

@@ -94,7 +94,6 @@ const Dashboard: React.FC = () => {
     setCurrentData(weeklyData);
   }
 
-
   const useLocationCheckin = () => {
     Alert.alert(
       'At the office',

maestro/Hourly.py

@@ -0,0 +1,151 @@
+import pandas as pd
+from sklearn.model_selection import train_test_split
+from sklearn.preprocessing import LabelEncoder, StandardScaler
+from tensorflow.keras.models import Sequential
+from tensorflow.keras.layers import Conv1D, MaxPooling1D, Flatten, Dense, Dropout, LSTM
+from tensorflow import keras
+import numpy as np
+import joblib
+
+# Load the Excel file
+file_path = 'datasets/Hourly_Predictions.xlsx'  # Replace with your actual file path
+excel_data = pd.ExcelFile(file_path)
+
+# Combine all sheets into one DataFrame
+sheets_data = {sheet: excel_data.parse(sheet) for sheet in excel_data.sheet_names}
+combined_data = pd.concat(sheets_data.values(), keys=sheets_data.keys(), names=['Day', 'Index']).reset_index(level=1, drop=True).reset_index()
+
+# Data Preparation
+label_encoder_day = LabelEncoder()
+combined_data['Day_Encoded'] = label_encoder_day.fit_transform(combined_data['Day'].fillna('Unknown'))
+
+# Extract hour from 'Time' column
+combined_data['Hour'] = combined_data['Time'].apply(lambda t: t.hour if pd.notnull(t) else 0)
+
+# Define bins and labels for each day (customized for each day of the week)
+day_bins = {
+    'Monday': [0, 50, 100, 150, 200, 250, 300, float('inf')],
+    'Tuesday': [0, 300, 600, 900, 1200, 1500, 1800, float('inf')],
+    'Wednesday': [0, 50, 100, 150, 200, 250, 300, float('inf')],
+    'Thursday': [0, 300, 600, 900, 1200, 1500, 1800, float('inf')],
+    'Friday': [0, 50, 100, 150, 200, 250, 300, float('inf')],
+    'Saturday': [0, 25, 50, 75, 100, 125, 150, float('inf')],
+    'Sunday': [0, 10, 20, 30, 40, 50, 60, float('inf')],
+}
+
+day_labels = {
+    'Monday': list(range(len(day_bins['Monday']) - 1)),
+    'Tuesday': list(range(len(day_bins['Tuesday']) - 1)),
+    'Wednesday': list(range(len(day_bins['Wednesday']) - 1)),
+    'Thursday': list(range(len(day_bins['Thursday']) - 1)),
+    'Friday': list(range(len(day_bins['Friday']) - 1)),
+    'Saturday': list(range(len(day_bins['Saturday']) - 1)),
+    'Sunday': list(range(len(day_bins['Sunday']) - 1)),
+}
+
+# Function to apply appropriate bins for each day
+def apply_bins(row):
+    day = row['Day']
+    bins = day_bins.get(day, [0, float('inf')])  # Default if not found
+    labels = day_labels.get(day, [0])
+    return pd.cut([row['Attendance']], bins=bins, labels=labels, include_lowest=True)[0]
+
+combined_data['Attendance_Bin'] = combined_data.apply(apply_bins, axis=1)
+
+# Drop missing bins and convert to integers
+combined_data.dropna(subset=['Attendance_Bin'], inplace=True)
+combined_data['Attendance_Bin'] = combined_data['Attendance_Bin'].astype(int)
+
+# Encode attendance bins
+y = combined_data['Attendance_Bin']
+
+# Standardize the 'Hour' feature
+scaler = StandardScaler()
+hour_scaled = scaler.fit_transform(combined_data['Hour'].values.reshape(-1, 1))
+joblib.dump(scaler, 'hourly_scaler.pkl')
+
+# Prepare features
+X_scaled = np.hstack([combined_data['Day_Encoded'].values.reshape(-1, 1), hour_scaled])
+
+# Reshape for CNN input
+X_scaled = X_scaled.reshape(X_scaled.shape[0], X_scaled.shape[1], 1)
+
+# Convert labels to categorical
+y_categorical = keras.utils.to_categorical(y)
+num_classes = y_categorical.shape[1]
+
+# Train-test split
+X_train, X_test, y_train, y_test = train_test_split(X_scaled, y_categorical, test_size=0.2, random_state=42)
+
+# Define CNN + LSTM model
+model = keras.Sequential([
+    keras.layers.Conv1D(input_shape=(X_train.shape[1], 1), filters=64, kernel_size=1, activation='relu', name='Conv1D'),
+    keras.layers.MaxPooling1D(pool_size=1, name='MaxPool1D'),
+    keras.layers.LSTM(50, return_sequences=False, name='LSTM'),
+    keras.layers.Flatten(name='Flatten'),
+    keras.layers.Dense(50, activation='relu', name='Dense_1'),
+    keras.layers.Dropout(0.5, name='Dropout'),
+    keras.layers.Dense(num_classes, activation='softmax', name='Output')
+])
+
+# Print the model summary
+model.summary()
+
+# Compile the model
+model.compile(optimizer='adam', 
+              loss=keras.losses.CategoricalCrossentropy(from_logits=False),  # From logits is False since output is softmax
+              metrics=[keras.metrics.CategoricalAccuracy()])
+
+# Define training parameters
+epochs = 30
+batch_size = 16
+testing = False
+
+# Train the model
+model.fit(X_train, y_train, epochs=epochs, batch_size=batch_size, validation_split=0.2)
+
+# Evaluate the model
+test_loss, test_acc = model.evaluate(X_test, y_test)
+print('\nTest accuracy: {}'.format(test_acc))
+
+# Save the model
+# model.export('python-code/models/hourly_predictions/1')
+
+def predict_for_day_hour(model, day, hour, label_encoder, scaler):
+    """
+    Predict attendance for a specific day and hour.
+
+    Parameters:
+    model: Trained Keras model used for predictions.
+    day: String representing the day of the week (e.g., 'Monday').
+    hour: Integer representing the hour of the day (0-23).
+    label_encoder: LabelEncoder instance used to encode day labels during training.
+    scaler: StandardScaler instance used to scale hour feature during training.
+
+    Returns:
+    predicted_bin: Predicted attendance bin for the input day and hour.
+    """
+    # Encode the day using the label encoder (same as training)
+    day_encoded = label_encoder.transform([day])[0]
+
+    # Scale the hour using the same scaler as in training
+    hour_scaled = scaler.transform(np.array([[hour]]))
+
+    # Prepare the input (reshape to match model's input shape)
+    input_data = np.array([[day_encoded, hour_scaled[0][0]]]).reshape(1, 2, 1)
+
+    # Predict the attendance bin (returns a probability for each class)
+    prediction = model.predict(input_data)
+    print(prediction)
+
+    # Convert the prediction from probabilities to class (attendance bin)
+    predicted_bin = np.argmax(prediction, axis=1)[0]
+
+    return predicted_bin
+
+# Example usage
+day_input = 'Monday'
+hour_input = 10  # 10 AM
+predicted_bin = predict_for_day_hour(model, day_input, hour_input, label_encoder_day, scaler)
+print(f"Predicted attendance bin for {day_input} at {hour_input}:00 is {predicted_bin}")
+

maestro/warmup_data(2).py

@@ -0,0 +1,51 @@
+# import tensorflow as tf
+# import numpy as np
+# import os
+# from tensorflow_serving.apis import predict_pb2, prediction_log_pb2
+
+# # Function to create representative input data
+# def create_warmup_data(input_shape):
+#     # Generate random data with the given input shape
+#     example_input = np.random.rand(*input_shape).astype(np.float32)
+#     return example_input
+
+# # Save the warmup data as a TFRecord in a format compatible with TensorFlow Serving
+# def save_warmup_data(file_path, warmup_data):
+#     # Create the PredictRequest for TensorFlow Serving
+#     warmup_request = predict_pb2.PredictRequest()
+#     warmup_request.model_spec.name = 'hourly_model'  # Replace with your model name
+#     warmup_request.model_spec.signature_name = 'serving_default'
+
+#     # Fill in the input data - ensure that 'input_data' matches your model's input signature name
+#     warmup_request.inputs['input_data'].CopyFrom(
+#         tf.make_tensor_proto(warmup_data, dtype=tf.float32)
+#     )
+
+#     # Create a PredictionLog containing the request
+#     warmup_log = prediction_log_pb2.PredictionLog(
+#         predict_log=prediction_log_pb2.PredictLog(request=warmup_request)
+#     )
+
+#     # Write to the TFRecord file
+#     with tf.io.TFRecordWriter(file_path) as writer:
+#         writer.write(warmup_log.SerializeToString())
+
+# # Model's input shape (batch size is None, so replace it with 1)
+# input_shape = (1, 2, 1)  # Replace with your model's input shape without the batch size
+
+# # Generate warmup data
+# warmup_data = create_warmup_data(input_shape)
+
+# # Define the file path for the warmup file
+# warmup_file_path = 'C:/Users/retha/Capstone/occupi/models/hourly_predictions/1/assets.extra/tf_serving_warmup_requests'
+
+# # Create the necessary directory if it doesn't exist
+# os.makedirs(os.path.dirname(warmup_file_path), exist_ok=True)
+
+# # Save the warmup data in TensorFlow Serving format
+# save_warmup_data(warmup_file_path, warmup_data)
+
+# print(f"Warm-up request saved to {warmup_file_path}")
+
+import tensorflow as tf
+print(tf.__version__)